INSTITUTE OF PHYSICS PUBLISHING
PHYSICS IN MEDICINE AND BIOLOGY
Phys. Med. Biol. 47 (2002) R85–R106
PII: S0031-9155(02)21840-0
TOPICAL REVIEW
Inorganic scintillators in medical imaging
Carel W E van Eijk
Delft University of Technology, IRI, Radiation Technology, Mekelweg 15, 2629 JB Delft,
The Netherlands
E-mail: vaneijk@iri.tudelft.nl
Received 4 January 2002
Published 5 April 2002
Online at stacks.iop.org/PMB/47/R85
Abstract
A review of medical diagnostic imaging methods utilizing x-rays or gamma
rays and the application and development of inorganic scintillators is presented.
1. Introduction
Inorganic scintillators are employed in most of the current medical diagnostic imaging
modalities using x-rays or gamma rays (see e.g. Webb (1990)). This is explained by the
comparatively good detection efficiency of inorganic scintillators for hard radiation. Yet,
the various diagnostic methods differ considerably and consequently the radiation detector
requirements also differ. These requirements are not always met by the scintillator
specifications. This is one of the reasons for continuous scintillator R&D. The other reasons
are the need for (1) new, digital diagnostic systems with an excellent image quality and a short
image acquisition time and (2) excellent quality real-time imaging systems for interventional
radiology at a low radiation dose. After a discussion of the various diagnostic methods in
which the emphasis is put on the expanding field of positron emission tomography, a review
of the specifications of the employed scintillators and some recent scintillator research is
presented.
2. X-ray imaging
In x-ray radiography (static imaging) an attenuation profile of a part of the human body is
projected onto a two-dimensional position-sensitive radiation detector (PSD) using the focus
of an x-ray tube as a point source. Thus, information on anatomical detail is obtained. The
method is applied in almost all fields of clinical practice. A variety of radiation detectors
are used. For example, at present, for chest radiography x-ray phosphor-screen-film cassettes
and storage-phosphor plates (computed radiography), both typically of 35 × 43 cm2, are
applied as the PSD on a large scale (Webb 1990). In most cases Gd2O2S:Tb is used as
the x-ray intensifying screen phosphor, i.e. the scintillator (Blasse and Grabmaier 1994).
0031-9155/02/080085+22$30.00
© 2002 IOP Publishing Ltd Printed in the UK
R85
R86
Topical Review
Figure 1. Schematic of a flat-panel detector. An array of amorphous-silicon diodes positioned on
a glass plate is covered by a scintillator screen of columnar grown CsI:Tl. Addressing lines and
read-out lines are respectively coming from and going to chips at the edge of the plate (courtesy
Philips Research Laboratories, Aachen).
The standard storage-phosphor material is BaFBr1−xIx:Eu (x 0.2) (Fuji, Agfa, Eastman
Kodak1) (Sonoda et al 1983). Furthermore, a digital chest-radiography system is used based on
the electrostatic read-out of a drum covered with an amorphous-selenium (a-Se) semiconductor
layer as the PSD (Thoravision, Philips) (Webb 1990, Neitzel et al 1994). Recently, new digital
radiography (DR) systems have been introduced; for example, Eastman Kodak developed a
system employing an a-Se photoconductor layer deposited on top of an amorphous-silicon
(a-Si:H) thin-film transistor array (Direct View). For the principle of detector operation, see
Soltani et al (1999). The active image area is 35 × 43 cm2, the pixel pitch is 139 µm. Agfa
introduced a system employing a Gd2O2S:Tb phosphor screen deposited on top of an array of
a-Si:H photodiodes coupled to an array of thin-film transistors, active area 43 × 43 cm2, pixel
size 160 × 160 µm2 (DR-Thorax).
A PSD such as the a-Se/a-Si:H system, which converts x-rays into electrons, is called
direct, as opposed to a PSD that makes use of the conversion of x-rays into light and
subsequently of light into electrons, such as the Gd2O2S:Tb/a-Si:H system, which is called
indirect. It appears that indirect flat-panel PSDs analogous to that of Agfa but employing
a CsI:Tl scintillator screen will find large scale application in the very near future. Canon,
General Electric, Varian and TRIXELL (a joint venture of Philips, Siemens and Thomson)
are following this path. TRIXELL is developing a 43 × 43 cm2 flat-panel PSD with a pixel
pitch of 143 µm (Pixium 4600). For the principle of operation, see figure 1 (Jing et al 1994,
Schiebel et al 1994) and also Moy (1999).
In mammography film-screen cassettes of 18 × 24 cm2 are generally used. In this case
a resolution of 0.1 mm is required to observe microcalcifications. Conventional film-screen
systems usually consist of two equal phosphor-binder layers (thickness up to ∼0.3 mm per
layer) and two equal emulsion layers, one on each side of the film base. More advanced
systems for chest radiography are asymmetric in structure, i.e. the two screens have different
layer thicknesses and the two emulsions are of different contrast (InSight, Eastman Kodak)
1
For company-related information the reader is referred to company websites.
Topical Review
R87
Image intensifier
phosphor screen
CsI:Na
columnar growth
focusing electrodes
anode 30 kV
X-rays
lens
electrons
CCD
camera
vacuum
output phosphor screen
photocathode
Figure 2. Schematic of an image intensifier system.
(Gray et al 1993). In mammography only one relatively thin screen (∼0.07 mm) and one
emulsion are used to realize the required resolution. As for efficiency, it should be noted that
relatively low-energy x-rays, e.g. produced at a tube voltage of 28 kV in a molybdenum anode
and filtered by 0.03 mm magnesium, are used for mammography compared to higher energy
x-rays, for example, produced at a tube voltage of 150 kV in a tungsten anode and filtered
by several millimetres of aluminium and also often by a few tenths of millimetres of copper,
for chest radiography. Fischer Imaging Corporation and TREX Medical have introduced DR
systems based on a phosphor screen coupled to charge-coupled devices (CCDs) by means
of fibre optical reducers. General Electric developed a digital mammography system based
on the CsI:Tl a-Si:H concept, 19 × 23 cm2, pixel size 100 × 100 µm2 (Senographe 2000D)
(Muller 1999).
In dentistry, where very small detectors of ∼2 × 3 cm2 to 3 × 4 cm2 are used for intra-oral
radiography (∼65 kVp x-rays), direct radiation detection in a CCD is applied, in addition
to the traditional intra-oral film-package and the storage-phosphor plate (van der Stelt 2001).
Pixel pitches are in the range 20–40 µm. Similar methods are applied for panoramic and
cephalometric radiography.
For fluoroscopy (dynamic imaging) a system is needed which responds fast enough to
provide real-time images of motion in the region of interest of the human body. At present
the x-ray image-intensifier-based system is the workhorse (Webb 1990, Hell et al 2000). A
CsI:Na scintillator screen on the inside of the entrance window absorbs the x-rays; see figure 2.
Scintillation photons are converted into electrons by means of a photocathode (e.g. Cs3Sb). The
electrons are accelerated and projected onto an output phosphor (e.g. ZnS:Cu, (Zn, Cd)S:Cu,
(Zn, Cd)S:Ag), resulting in an intermediate, reduced image. In the early days the light thus
produced was observed by means of a TV camera. Today the intermediate image is observed
by means of advanced CCD systems. Depending on the application, image intensifier screen
diameters are used in the range 23–40 cm, providing 1 k × 1 k pixel images, at frame rates up
to 30 frames per second.
Image-intensifier-based systems are also widely used for radiographic imaging and
for simulation in radiation therapy planning (radiographic and fluoroscopic) (Williams and
Thwaites 2000).
The image intensifier systems are very bulky and research to replace these detectors by
much more practicable devices is in progress. Most interesting is the development of large
∼40 × 40 cm2 fluoroscopic systems based on the CsI:Tl/a-Si:H flat-panel concept.
R88
Topical Review
X-ray fan beam
(rotating)
+
1D position
sensitive
detector
(rotating)
X-ray source
(rotating)
Centre of rotation
Figure 3. Schematic of an x-ray CT system.
Summarizing, we note that, apart from the a-Se-semiconductor-based PSDs and the dental
CCDs, all systems are luminescence based. The screen phosphor Gd2O2S:Tb, scintillators
CsI:Na and Cs:Tl, and storage-phosphor BaFBr1−xIx:Eu are dominating the x-ray imaging
field. We will return to these devices in section 6.
3. X-ray computed tomography
In principle, in x-ray computed tomography (CT) the body is consecutively irradiated from a
large number of directions by an x-ray fan beam (Kalender 2000); see figure 3. Attenuation
profiles are registered and from them cross-sectional images of the body are reconstructed.
Actually, sophisticated scan modes are used such as multi-slice spiral CT combined with
volume reconstruction. Applications are found in many fields of clinical practice.
To measure the profiles, a segment of a circle, radius ∼1 m, centre on the x-ray source, is
covered with up to ∼1000 small detectors which form a one-dimensional PSD. The segment
rotates with the x-ray source (tube) for detecting the fan beam at all angular settings (figure 3).
X-ray energies are in the range up to ∼140 keV (80–140 kVp) and consequently an inorganic
scintillator is the obvious detection medium. CdWO4, Bi4Ge3O12 (BGO) and CsI:Tl, read out
by photodiodes (integrating mode), have been used in the past. However, for various reasons
these scintillators are less favourable; for example, light yields are low or/and matching
with the quantum efficiency curve of the light detecting diodes is poor (CdWO4, BGO).
Furthermore, at each angular position a certain amount of light is accepted due to the slow
scintillation response on radiation absorbed at the previous angular position (primary decay
and afterglow). In a modern CT system the rotation time is typically <1 s. A preferred
condition is, for example, that <0.01% of the maximum scintillation intensity is emitted after
∼3 ms (Blasse and Grabmaier 1994, Hupke and Doubrava 1999). The afterglow of CsI:Tl
is clearly too high (table 1). Another undesirable effect is radiation damage, resulting in a
change of light output, e.g., an increase in CsI:Tl and a decrease in CdWO4. Furthermore,
CdWO4 is preferably avoided because of its toxicity.
One way to remedy the shortcomings has been to employ pressurized Xe gas ionization
detectors; for example, in one of the Siemens SOMATOM systems we encounter an arc of 768
xenon sensors of 1.2 × 18 mm2 and a depth of several centimetres. However, the detection
efficiency is not high enough.
A very interesting development was the preparation of scintillators using advanced ceramic
technologies, and the introduction of co-doping to reduce the afterglow (Grabmaier and
Rossner 1993, Blasse and Grabmaier 1994, Greskovich and Duclos 1997). Examples are
Density
(g cm−3)
ρZ4eff
(106)
Attenuation length
at 511 keV (mm)/
prob. phot. eff.
(%)
Hygroscopicity
Light yield
(photons/
MeV)
CsI:Na
CsI:Tl
4.51
4.51
38
38
22.9/21
22.9/21
Yes
Slightly
40 000
66 000
CaWO4
YTaO4:Nb
Gd2O2S:Tb
Gd2O2S:Pr,Ce,F
Gd2O2S:Pr (UFC)
Y1.34,Gd0.60O3:(Eu,Pr)0.06c
(Hilight)
Gd3Ga5O12:Cr,Ce
CdWO4
Lu2O3:Eu,Tb
CaHfO3:Ce
SrHfO3:Ce
BaHfO3:Ce
NaI:Tl
LaCl3:Ce
LaBr3:Ce
Bi4Ge3O12 (BGO)
Lu2SiO5:Ce (LSO)
Gd2SiO5:Ce (GSO)
YAlO3:Ce (YAP)
LuAlO3:Ce (LuAP)
Lu2Si2O7:Ce (LPS)
6.1
7.5
7.3
7.3
7.3
5.9
89
96
103
103
103
44
13.6/32
11.8/29
12.7/27
12.7/27
12.7/27
17.8/16
No
No
No
No
No
No
20 000b
40 000b
60 000b
35 000b
50 000b
42 000b
7.1
7.9
9.4
7.5
7.7
8.4
3.67
3.86
5.3
7.1
7.4
6.7
5.5
8.3
6.2
58
134
211
139
122
142
24.5
23.2
25.6
227
143
84
7
148
103
14.8/18
11.1/29
8.7/35
11.6/30
11.5/28
10.6/30
29.1/17
27.8/14
21.3/13
10.4/40
11.4/32
14.1/25
21.3/4.2
10.5/30
14.1/29
No
No
No
No
No
No
Yes
Yes
Yes
No
No
No
No
No
No
40 000b
20 000b
30 000b
∼10 000b
∼20 000b
∼10 000b
41 000
46 000
61 000
9 000
26 000
8 000
21 000
12 000
30 000
E/E (FWHM)
at 662 keV (%)
PMT read-out
Decay time
(ns)
Afterglow
(% after
3 ms/100 ms)
Emission
maximum
(nm)
630
800–>6 × 103
>2/0.3
420
550
<0.1/<0.01
0.02/0.002
4.9/<0.01
420
410
545
510
510
610
7.4
6.6 (PMT)/
4.3 (SDD)a
Integrating mode
Integrating mode
Integrating mode
Integrating mode
Integrating mode
Integrating mode
730
495
611
390
390
400
410
330
358
480
420
440
350
365
380
Integrating mode
6.8
Integrating mode
Integrating mode
Integrating mode
Integrating mode
5.6
3.3
2.9
9.0
7.9
7.8
4.3
∼15
∼10
1 × 106
4 × 103
3 × 103
1 × 106
140 × 103
5 × 103
>106
40
40
25
230
25 (65%)
35 (90%)
300
40
60
30
18
30
<0.1/0.01
<0.1/0.02
>1/0.3
t (ps)
/BaF2
/60Co
Topical Review
Table 1. Inorganic scintillators for medical imaging (for references see text).
224
385
a
Measured with a PMT and silicon drift detector (SDD).
Measured at ∼60–80 keV; all others at 662 keV.
c Proprietary co-dopant.
b
R89
R90
Topical Review
Figure 4. Schematics of GE curved x-ray CT detectors. One-dimensional PSD on the left, and
two-dimensional array detector for multi-slice system on the right (from McCollough and Zink
(1999), courtesy the authors and American Association of Physicists in Medicine).
(Y,Gd)2O3:Eu3+ co-doped with Pr and Gd2O2S:Pr co-doped with Ce and F. In about 1998
several companies (General Electric, Marconi, Siemens, Toshiba) introduced two-dimensional
array-detector arcs employing these types of ceramic scintillators; see figure 4. Thus multislice (spiral) imaging became possible (see e.g. McCollough and Zink (1999)). We will return
to the ceramic scintillators in section 6.
4. Radionuclide imaging—single-photon emission
In radionuclide imaging based on single-photon emission, the gamma radiation emitted by
a radiopharmaceutical introduced into the body is used either for simple projections (planar
scintigraphy) or for single-photon emission computed tomography (SPECT) (Webb 1990). In
both cases gamma cameras are used, originally typically circular with a diameter up to 50 cm,
employing a monolithic NaI:Tl crystal plate (thickness ∼6–12 mm) and photomultiplier
tube (PMT) read-out (e.g. hexagonal packing of 61 PMTs) (Short 1984, Jasczak et al 1980).
Modern systems are square or rectangular, up to ∼50 × 60 cm2, with the thickness of the
NaI:Tl crystal plate up to 25 mm. In principle the photon-energy range is 400 keV. However,
140 keV from 99mTc labelled compounds is used most frequently. In scintigraphy large heavy
collimator plates are used with parallel channels (diameter ∼2–3 mm) and 1:1 projection
images of the distributions of the radiopharmaceuticals are obtained. In SPECT cone beam
Topical Review
R91
Figure 5. Hybrid SPECT/PET system IRIX with three gamma cameras (courtesy Philips/Marconi
Medical Systems).
collimation is used. Projections are measured from different sides of the body from which
three-dimensional images are reconstructed.
The gamma camera operates in counting mode. For each event the distribution of the
scintillation light over the PMTs is used to obtain the point of interaction. In general the
information is extracted by means of a resistor network, connecting the PMTs (Anger logic).
For SPECT (and PET, section 5) cameras with signal digitization per PMT are also used
(e.g. ADAC/UGM). Recently Siemens/CTI introduced an experimental gamma camera for a
hybrid SPECT/PET system, operating with arrays of small NaI:Tl and Lu2SiO5:Ce (LSO:Ce)
crystals read out by PMTs (Jones et al 2001); see also the next section. In figure 5 an example
of a hybrid SPECT/PET system with three gamma cameras operating in SPECT mode is
shown.
5. Radionuclide imaging—positron emission tomography
5.1. Introduction
Positron emission tomography (PET) is a very powerful medical diagnostic method to observe
the metabolism, blood flow, neurotransmission and handling of important biochemical entities
(Webb 1990). PET is typically applied in neurology, oncology and cardiology. Imaging
is realized by means of two 511 keV quanta which are emitted approximately collinearly
when a positron, emitted by a radiopharmaceutical introduced into a patient, annihilates in
tissue. The two annihilation quanta are detected position sensitively in coincidence (see
figure 6). To a good approximation, the point of positron emission is situated on the line
of response connecting the two detection positions. Many annihilations give many lines of
R92
Topical Review
Collinearly emitted
annihilation quanta
detected in coincidence
Retracted
BGO + PMT
detector rings
(inner diam. ~ 0.8 m)
Septa
Figure 6. Schematic of PET system.
BGO detector block
8 x 8 columns
of 6 x 6 x 30 mm3
A
B
4 PMTs
C
D
y
y=
x
30 mm
x=
(A+B) – (C+D)
A+B+C+D
(B+D) – (A+C)
A+B+C+D
Figure 7. Schematic of BGO detector block showing saw cuts of 8 × 8 columns (left), to be
coupled to four PMTs for position-sensitive light detection (right).
response (LOR) and in principle the three-dimensional reconstruction of the LOR represents
the radiopharmaceutical distribution, i.e. the image.
For position-sensitive detection, a PET system consists, in general, of many rings with
thousands of scintillation detectors (ECAT systems, Siemens/CTI) (see figures 6 and 7).
BGO is the most used scintillator because of its good detection efficiency and large probability
of photoelectric effect (40%) for 511 keV quanta (table 1). Yet, other PSD systems are
also applied. ADAC/UGM employed six curved NaI:Tl crystals (2.54 cm thick) to form a
cylinder, ∼25 cm high, ∼90 cm in diameter, read out by PMTs (CPET) (Adam et al 2001).
NaI:Tl has the advantage that large, curved crystals can be produced relatively easy. Thus,
acquisition over a large solid angle is realized. However, the intrinsic detection efficiency
of NaI:Tl is significantly smaller than that of BGO (see table 1). Furthermore, we come
across systems consisting of two gamma cameras (without collimator plates) facing each other
(gamma camera PET, GC-PET). An example is the Millenium VG system of General Electric,
Topical Review
R93
utilizing cameras with a 16 mm thick NaI:Tl crystal, a field of view of 54 × 40 cm2, 59 PMTs
and one ADC per PMT. Systems with 25 mm thick NaI:Tl crystals are also available.
There is much interest in introducing new scintillator materials. Other properties of
BGO, such as its light yield, energy resolution and response time, make it less than optimal
for application in PET. Siemens/CTI developed a new system based on LSO:Ce. In the ECAT
Accel system, the BGO crystals used in the ECAT EXACT have been replaced by 25 mm
LSO:Ce crystals (Townsend 2001). Furthermore, Siemens/CTI introduced the LSO:Ce-based
high resolution research tomograph (HRRT) for brain studies (Wienhard et al 2001) and
ADAC/UGM developed a brain PET system utilizing Gd2SiO5:Ce (GSO:Ce), the ALLEGRO
(Karp et al 2001). In section 4 the hybrid SPECT/PET system based on NaI:Tl/LSO:Ce
gamma cameras (Jones et al 2001) has been mentioned.
In addition to company-related R&D, many groups are actively developing new PET
methods and techniques; as an example, we mention the use of liquid xenon (LXe) as a
scintillator and ionization-chamber material (Chepel et al 1997, Collot et al 2000). Many
papers on PET can be found in the proceedings of recent IEEE Nuclear Science Symposiums
& Medical Imaging Conferences (NSS/MIC 1999, 2000).
For the sake of completeness it should be mentioned that various application-specific
dedicated PET systems are being developed; for example, there is serious interest in application
of PET in combination with mammography (positron emission mammography) (e.g. see Lecoq
2001). LSO:Ce is considered for utilization. Furthermore, animal studies can be performed
for the development of radiopharmaceuticals for human use and for gene expression imaging
and post-genomic investigations. Consequently, animal-PET systems are needed. Concorde
introduced an LSO:Ce-based small-animal PET system (MicroPET). Several other smallanimal PET systems have been developed (see e.g. Weber et al (2000) and Del Guerra et al
(2000)). In both cases the relatively low density, small-Z scintillator YAlO3:Ce (YAP:Ce) is
employed.
It should be noted that multi-modality systems are becoming increasingly important.
Combined CT–PET systems have been introduced by ADAC/Philips (Gemini), General
Electric (Discovery series) and Siemens/CTI (Biograph) (Beyer et al 2000).
The properties of the PET scintillators are discussed in section 6. However, first PETsystem requirements will be addressed somewhat more in detail.
5.2. PET-system requirements and sources of error
5.2.1. Detection efficiency. The detection efficiency for a 511 keV quantum (ε) is found
squared in the coincidence efficiency. Consequently, it is very important to make ε as high as
possible. This implies that both the solid angle subtended by the PET system and the intrinsic
detector efficiency should be large. Only then can images be realized with a limited amount
of radioactive material in a reasonable time. This explains why many detector rings are used
and BGO is most often employed as the scintillator material.
In general BGO blocks with saw-cuts are applied (figure 7). Depending on the system
these cuts provide 36 to 64 crystal columns of, e.g., 4 × 4, 3 × 6, 6 × 6 or 4 × 8 mm2 cross
sections, coupled at the base in such a way that the scintillation-light distribution allows the
determination of the column hit by a radiation quantum using four PMTs or two PMTs
with a dual structure, and Anger-type logic (Casey and Nutt 1986, DeGrado et al 1994,
Wienhard et al 1994). Detector blocks are combined in rings to form typically 24–48 planes
of scintillator columns. Using coincidences between scintillator columns in the same plane
and cross-plane coincidences between columns in adjacent planes, image slices are obtained
at a pitch of half the plane distance.
R94
Topical Review
Originally, planes were separated by septa, i.e. lead collimator plates. In an actual PET
scan a significant fraction (0.3–0.5) of the 511 keV quanta are Compton scattered in the
patient. This gives a continuous background. The septa reduce this background significantly.
With septa, cross-plane 511–511 keV coincidences are limited to differences of 3–4 planes
(two-dimensional imaging). Modern systems can be operated with removed (retracted) septa
where coincidences between many planes are accepted resulting in three-dimensional (3D)
imaging (Townsend et al 1989, Cherry et al 1991). The CPET has no septa at all (Adam et al
2001). With septa retracted, the increased solid angle and consequently a higher 511–511 keV
coincidence rate compensates to a certain extent for the increased Compton-scattering
background. This background is reduced only by accepting events in the full-energy peak of
the pulse-height spectrum (the better the resolution, the better the reduction will be; typical
energy resolutions in PET systems are: for BGO 20% FWHM, for NaI:Tl ∼10% FWHM).
However, many events with 511 keV quanta interacting in a scintillator due to the Compton
effect will be eliminated as well. Consequently, depending on the PET system and specific
application, one has to decide whether a full-energy peak window is used or a more relaxed
window.
In a small-animal PET system the Compton scattering in a small animal plays a minor role.
Consequently, Compton events in the pulse height spectrum are primarily due to scattering in
the scintillator and can be accepted. Therefore, the threshold can be decreased. However, as
a result position resolution may become worse as the Compton-scattered quantum may give a
signal above threshold in a neighbouring crystal.
As for the scintillator one should realize that absorption by photoelectric effect per cm
is proportional to ρZ 3–4 with ρ the density and Z the effective atomic number. For 511 keV
quanta the chance of photoelectric effect at Z = 80 is no more than 50%, the other 50% being
the Compton effect (proportional to ρ). Depending on the kind of scintillator and crystal size,
detection of both the Compton electron and the Compton-scattered photon in the same crystal
may significantly contribute to the intensity of the full-energy peak.
5.2.2. Random coincidences and dead time. Random coincidences will occur, e.g. , if within
the coincidence-time window one 511 keV quantum of each of the two pairs of annihilation
quanta is detected and the other is emitted or Compton scattered by the patient in a direction
outside the PET system. Random coincidences show up as a continuous background
in an image. The ratio of true to random coincidences is inversely proportional to the
coincidence-time resolution 2τ . The time resolution depends strongly on the response time
of the scintillator, its light yield and light-detection efficiency. Using BGO, 2τ is in the order
of 10 ns. A much smaller 2τ will reduce the background significantly. Furthermore, if 2τ is
very small, time-of-flight information can be used to select a region of interest; for example,
at 2τ = 1 ns, the diameter of such a region will be 15 cm.
The true-to-random coincidence ratio is also affected by energy selection (Muehllehner
et al 2001). The random coincidence intensity is approximately proportional to the square
of the singles-spectrum intensity. A system without septa has considerably more Comptonscattered events in the singles spectrum than in the true coincidence spectrum. Consequently,
in the random coincidence spectrum there are relatively more events in the region of the
Compton continuum than in the photopeak in comparison with the corresponding parts of the
true coincidence spectrum. Therefore, selection of photopeak events will improve the true-torandom coincidence ratio. A good energy resolution and a high full-energy peak intensity are
indeed important.
Another important aspect is dead time, i.e. the time in which a coincidence cannot be
registered because the PET system is too busy handling a previous coincidence event. Several
Topical Review
R95
parts of the system contribute to the dead time and the detector is one of them. Taking the
BGO response time of 300 ns (scintillation decay time) as the dead time, we would lose about
3% of the events at a rate of 105 per second per read-out channel, which is approximately
the maximum rate of a 64 scintillator–column block in a modern PET system. However, it
should be noted that the actual dead time of a block is at the microsecond level due to the
signal processing (shaping times). If the scintillator columns were read out individually, the
rate would be only ∼1000 per second and the dead time of the scintillator/detector including
electronics would be negligible.
5.2.3. Position resolution. On emission a positron travels a distance of the order of 1 mm
before it is thermalized and annihilates with an electron. This electron itself has some
momentum resulting in a small deviation of the angle between the annihilation quanta from
180◦. These two effects contribute an additional intrinsic position resolution of up to 2 mm
FWHM for a 0.8 m diameter system, dependent upon the energy, and therefore range in tissue,
of the emitted positron. The intrinsic resolution decreases with decreasing system diameter.
If they are not Compton scattered in the patient, annihilation quanta emitted from the
central axis of a PET system and in the plane of a detector ring will hit a scintillation
detector in a direction more or less parallel to the length of the crystal columns (figure 7).
Quanta emitted off-centre hit the detector at an angle. If it is not absorbed in the first crystal
column encountered, the 511 keV quantum will traverse it and hit a neighbouring column
well behind the detector entrance window. If it is detected there a parallax error arises. It is
known as radial elongation due to its manifestation upon image reconstruction. This effect
decreases with increasing system diameter, i.e. it leads to a requirement opposite to that in the
previous paragraph and the large-solid-angle condition. Furthermore, for parallax reduction
the length of the scintillator column should be as small as possible without loss of detection
efficiency. However, for BGO a relatively long crystal, typically 30 mm, is used for efficient
detection. This explains efforts to introduce methods that give depth-of-interaction (DOI)
information.
A large variety of DOI methods have been studied. The response of a scintillator (light
yield and decay time) is temperature dependent. This offers a possibility to extract DOI
information. It was studied for NaI:Tl and BGO (Karp and Daube-Witherspoon 1987).
In other studies columns of two or more different crystals in line were used for obtaining
different responses as a function of depth, e.g., LSO:Ce+GSO:Ce (Saoudi et al 1999, Saoudi
and Lecomte 1999) and LSO:Ce+GSO:Ce+BGO (Seidel et al 1999). In the HRRT system
columns of two 7.5 mm LSO:Ce crystals with different decay times are used (Wienhard et al
2001). Recently, a method was presented employing avalanche photodiodes (APDs) on both
sides of a 22 mm long LSO:Ce column (Shah 2001). A DOI resolution of ∼3 mm was
obtained. For the many methods studied or proposed to obtain DOI information, see also
Meng and Ramsden (2000).
Present PET systems have a position resolution of typically ∼4 mm FWHM in the centre
of the system, increasing to ∼5–6 mm if we move in the axial direction to the edge of the
system or in the radial direction to, say, 20 cm from the axis. The position resolution
we are aiming at in future systems is at the level of 2–3 mm. Then the cross section of the
crystal columns should be about 2 × 2 mm2 and DOI information will become even more
important.
As for light detection it should be mentioned that in addition to the PMTs employed in
most of the PET systems described in this section, by now there are several position-sensitive
light sensors available that are applicable for read out of individual scintillator columns,
namely, APD arrays, position-sensitive PMTs (e.g. used in the Concorde MicroPET system)
R96
Topical Review
and multi-pixel hybrid photomultipliers. The reader is referred to, e.g., Bourgeois et al (1999).
These new sensors will certainly find their way into the field of medical imaging.
6. Inorganic scintillator development
6.1. Inorganic scintillator basics
In developing new scintillator materials, on one hand, we will in general select a material
with a relatively high density ρ and high atomic number Z for efficient x-ray and gamma-ray
detection (absorption probability by photoelectric effect per cm ∝ ρZeff3–4). This material
should transmit scintillation light efficiently and consequently we rely on ionic crystals, or
crystals with some degree of covalence but with a forbidden-gap energy between valence and
conduction bands, Egap, large enough to transmit the emitted light. On the other hand, for
good energy, time and position resolution
we need a large number of scintillation photons,
Nph, (relative statistical spread ∝ 1/ Nph ), and consequently the forbidden gap should be as
small as possible (Rodnyi et al 1995):
E
Nph =
SQ.
(1)
βEgap
The first term on the right-hand side represents the number of thermalized electron–hole
pairs Ne–h produced by the absorption of gamma-ray energy E. The parameter β indicates the
average energy required to produce one thermalized electron–hole pair: Ee–h = βEgap, β =
∼2–3. S is the transport/transfer efficiency of the e–h pair/energy to the luminescence centre
(LC) of the scintillator, and Q is the quantum efficiency of the LC, i.e. the efficiency for photon
emission once the LC is excited. Of the three stages the transport/transfer, denoted by S, is
the least predictable. It depends very much on defects present in the scintillator, other than
the LC, that may capture electrons or holes or both, and produce radiationless transitions or
afterglow upon thermal excitation. These defects can arise from the interaction itself, from
crystal growing, or due to impurities; for more details see papers in Dorenbos and van Eijk
(1996), Yin Zhiwen et al (1997) and Mikhailin (2000).
Another aspect is non-proportionality. Most scintillators have a light yield, expressed in
photons per MeV of absorbed radiation energy, which is energy dependent (see e.g. Dorenbos
et al (1995) and van Eijk et al (2001)). This implies that for monoenergetic gamma rays the
full-energy peak resulting from photoelectric effect in a scintillator is not at the same position
in the pulse height spectrum as the peak resulting from absorption of a Compton electron
followed by absorption of the corresponding scattered Compton quantum. Consequently, the
energy resolution is significantly degraded compared with the result expected on statistical
grounds (see e.g. Valentine and Rooney (1997)). At present it is not possible to select materials
for which non-proportionality effects can be estimated a priori.
In table 1 specifications of several scintillator materials of interest for medical imaging are
4
as a measure of the probability of interaction by
summarized. Presented are the density, ρZeff
photoelectric effect, the attenuation length and probability of interaction by photoelectric effect
at 511 keV, hygroscopicity, light yield, decay time, for some materials afterglow information,
wavelength of the emission maximum, for some scintillators energy resolution at 662 keV
(FWHM) and time resolution using BaF2 as the second, coincidence scintillator.
Of the presented materials, CsI:Na, CsI:Tl, CdWO4, NaI:Tl and BGO are traditional
scintillators; for reference, see van Eijk (2001) and Dorenbos et al (1995). Applied as an
x-ray phosphor CsI:Na, CsI:Tl, CaWO4, YTaO4:Nb, Gd2O2S:Tb, Gd2O2S:Pr, Ce, F, Gd2O2S:Pr
(UFC), Y1.34,Gd0.60O3:(Eu, Pr)0.06 (Hilight) and Gd3Ga5O12:Cr,Ce are partly well known.
The developments will be discussed in the next section. In contrast to the scintillators just
Topical Review
R97
mentioned, all of which have a relatively slow scintillation response, the Ce-based scintillators
have a much faster response. The Ce3+ ion has one electron in the 4f state, which is lifted
to the empty 5d shell upon excitation. Subsequent de-excitation will occur by an allowed
5d → 4f electric dipole transition with a decay time in the order of 30 ns. These relatively new
materials will also be discussed in the next section; for reference, see sections 2, 3 and 6.2.
6.2. Inorganic scintillators for medical imaging
6.2.1. X-ray scintillators. The scintillator employed most frequently in x-ray intensifying
screens is Gd2O2S:Tb. The emission results from Tb3+ forbidden 4f → 4f transitions. The
highest intensity line is at 545 nm, i.e. in the green part of the spectrum. Examples of
x-ray phosphors with blue emissions are CaWO4, one of the earliest applied phosphors, and
YTaO4:Nb (see table 1). The emission of the former is due to the WO2−
4 complex that acts as
a luminescence centre. Undoped YTaO4 emits at ∼350 nm due to a charge-transfer transition
in the tantalate group. Doping with Nb shifts the emission to a longer wavelength (see e.g.
Blasse and Grabmaier (1994)).
In addition to the general detection efficiency aspects discussed in section 6.1, for x-ray
detection efficiency the position of the K-edge is very important. In the energy region between
the K-edges of Gd and W, which are, respectively at 50.2 and 69.5 keV, CaWO4 is less
sensitive to x-rays than Gd2O2S:Tb. YTaO4:Nb is more sensitive than Gd2O2S:Tb at energies
<50.2 keV and more sensitive than CaWO4 at energies <69.5 keV. Above 69.5 keV the
absorption coefficients are approximately the same. As in addition the light yields of
Gd2O2S:Tb and YTaO4:Nb are relatively high (table 1), both offer high-speed imaging using,
respectively, a green- and blue-sensitive radiographic film; for more details, see e.g. Miura
(1998).
Although not a scintillator, we mention BaFBr1−xIx:Eu (x 0.2) for completeness. This
storage-phosphor material was introduced in the early 1980s by Fuji (Sonoda et al 1983).
With it the computed radiography method was presented, employing a storage-phosphor plate
and photon-stimulated luminescence. In principle the operation of the storage phosphor is
clear. However, the exact mechanism is still unknown. The electrons produced by x-rays
are trapped in Br− vacancies. The nature of the hole trap centres is not known (Spaeth
2001, Lakshmanan et al 2001). Upon optical stimulation, the trapped electron is released and
electron–hole (exciton) recombination occurs under simultaneous excitation of Eu2+, followed
by luminescence. The addition of iodine results in a shift of both the optical stimulation
spectrum and the emission spectrum to longer wavelength.
Both storage-phosphor plates and x-ray phosphor screens are formed by the addition of
a binder to the phosphor powder. The ensemble is deposited on a screen base. Due to the
scattering of the stimulation light at the birefringent BaFBr crystallites in a storage-phosphor
plate, the position resolution is limited; for example, the resolution of ∼100 µm required
for mammography could not yet be realized. In x-ray phosphor screens there is significant
scintillation-light spreading by the scattering from the phosphor grains, in particular when the
screen thickness is increased for x-ray detection-efficiency reasons. A method to reduce this
is the application of columnar growth. This is possible with CsI:Na and CsI:Tl. Columnar
screens of CsI:Na have long been mounted inside image intensifiers. The emission, due to
an exciton bound to the Na+ ion, matches well with the photocathode sensitivity. With the
introduction of the a-Si:H flat-panel concept, CsI:Tl became the obvious columnar-phosphor
candidate. Firstly, hygroscopic CsI:Na operates well in the vacuum of the image intensifier
tube but will not do so in the flat-panel detector. CsI:Tl will not suffer from this problem as it
is hardly hygroscopic in comparison with CsI:Na. Secondly, the emission spectrum of CsI:Tl,
R98
Topical Review
1.0
0.8
0.6
a-Si:H
CsI:Tl
0.4
0.2
0.0
400 500 600 700 800
wavelength/nm
Figure 8. Matching of CsI:Tl emission spectrum with photon sensitivity of a-Si:H diodes (arbitrary
units along vertical scale; courtesy Philips Research Laboratories, Aachen).
peaking at 550 nm, matches the quantum efficiency curve of amorphous silicon quite well;
see figure 8.
From figure 9 an impression of the quality of columnar growth of CsI:Tl developed
in recent years is obtained. The lengths of the ∼3 µm diameter columns are well over
0.5 mm, giving an x-ray detection efficiency at, e.g., 60 keV of >83%. It should be noted
that the luminescence properties of these columns differ from those of CsI:Tl scintillation
crystals (Garnier et al 2000). In addition to the two standard emission bands at 400 nm
(Tl+ emission) and 550 nm (strongly perturbed thallium-bound exciton emission), at liquid
nitrogen temperature a band can be observed at 460 nm, which is attributed to a weakly
perturbed thallium-bound exciton centre. At higher Tl-doping concentrations this emission
disappears and the emission at 550 nm dominates. At ambient temperature the extra emission
band is not observed.
As already mentioned in section 3, CsI:Tl suffers from both afterglow and an increase in
light yield resulting from radiation damage (hysteresis). Recent studies resulted in a model
explaining both effects by deep trapping of x-ray-generated carriers (Wieckzorec and Overdick
2000). Furthermore, the afterglow could be reduced to a level low enough for flat-panel
application. Studies are being continued.
X-ray CT has flourished with the introduction of ceramic scintillators, in particular
Gd2O2S:Pr, Ce, F, Gd2O2S:Pr (UFC) and Y1.34,Gd0.60O3:(Eu, Pr)0.06, respectively, developed by
Hitachi Metals (Yamada et al 1989), Siemens (Rossner et al 1999, Hupke and Doubrava 1999)
(brand name UFC, ultra fast ceramics) and General Electric (Greskovich and Duclos 1997)
(brand name Hilight). These polycrystalline ceramic scintillators couple good scintillation
properties with homogeneity and good machinability. In addition, their emission wavelengths
match well with the sensitivity of silicon diodes.
In Gd2O2S, or GOS, based ceramics scintillation arises from 4f → 4f transitions in the Pr3+
dopant ion. The decay time of 3–4 µs does not pose a problem for the condition of <0.01%
light yield after 3 ms (section 3). The afterglow is due to the trapping of charge carriers that
are subsequently thermally released and recombine at luminescence centres. Co-doping with
Topical Review
R99
Figure 9. Vapour-deposited column-shaped CsI:Tl scintillation crystals of very smooth structure.
Diameter ∼3 µm, length >0.5 mm (courtesy Philips Research Laboratories, Aachen).
cerium and fluor results in non-radiative recombination and reduction of trap centres, thus
reducing the afterglow. However, the scintillation intensity is also reduced. Consequently, a
compromise has to be made between light yield and afterglow reduction. It should be noted
that co-doping of Gd2O2S:Pr with Tb and F leads to similar results (Nakamura and Ishii 2000).
In UFC the afterglow reduction is primarily realized by optimized ceramic processing. From
the afterglow values after 3 ms in table 1, we observe that only UFC is close to the <0.01%
condition. Also, for the second condition, the light yield after 100 ms (Hupke and Doubrava
1999), UFC has the smallest value.
Due to scintillation-light scattering at the hexagonal grain boundaries in GOS-based
ceramic materials these are translucent. Fortunately, relatively thin scintillation detectors can
4
values, table 1), i.e.
be used because of the high x-ray absorption efficiency (compare ρZeff
∼1.5 mm compared to ∼3 mm for transparent Hilight to absorb 98% of 120 keV x-rays
(Hupke and Doubrava 1999). Therefore, light collection does not pose a problem.
In Y1.34,Gd0.60O3:(Eu, Pr)0.06, also referred to as YGO, scintillation arises from 4f → 4f
transitions in the Eu3+ dopant ion. Pr3+ or Tb3+ co-dopant ions reduce the afterglow as they
act as hole traps in competition with intrinsic traps responsible for the afterglow. The Pr
or Tb sites decay non-radiatively (Greskovich and Duclos 1997). Considering the afterglow
properties in table 1, we observe a much higher afterglow after 3 ms than for GOS ceramics.
Hilight has the advantage of being transparent.
R100
Topical Review
Another interesting ceramic scintillator is Gd3Ga5O12:Cr, Ce (Greskovich and Duclos
1997). Scintillation arises from the Cr3+ luminescence centre and afterglow reduction
from Ce3+ co-doping, presumably by a process similar to that in YGO. Some new ceramic
materials are Lu2O3:Eu, Tb (Dujardin et al 2000, Lempicki et al 2001) and barium, strontium
and calcium hafnates doped with cerium (Greskovich and Duclos 1997, Venkataramani et al
2001). In principle Lu2O3:Eu, Tb and the hafnates are of interest for application in both CT and
PET because of their small attenuation lengths at 511 keV (table 1). However, for Lu2O3:Eu, Tb
the decay time and afterglow are too high. Unfortunately, doping of Lu2O3 with cerium
does not result in fast luminescence as the lowest 5d level appears to be positioned in the
conduction band (Happek et al 2000). Of the hafnates, SrHfO3:Ce has the highest light yield,
∼20 000 photons/MeV (table 1). This is somewhat small. The decay times of 25–40 ns,
characteristic of Ce3+ emission, are very attractive. Research on these new materials is in
progress.
6.2.2. Radionuclide imaging, positron emission tomography. NaI:Tl has long been the
standard scintillator of the gamma camera. The luminescence is from the Tl+ ion. The light
yield is high and the emission wavelength matches the sensitivity of PMTs quite well. Crystals
of large size (diameter ∼0.8 m, height ∼0.8 m) can be grown relatively easily. In principle the
hygroscopicity of NaI:Tl is a disadvantage. However, canning methods are well developed.
Recently, the new scintillators LaCl3:Ce and LaBr3:Ce have been introduced (van Loef
et al 2000, 2001). Their development may, among other things, be of interest for application
in the gamma camera. These materials have a high light yield, fast response (Ce3+ ion),
excellent energy resolution and very good time resolution. It should be noted that the energy
and time resolution presented in table 1 have been obtained with different experimental set-ups
optimized for their specific tasks (see next paragraph). The detection efficiency is comparable
to that of NaI:Tl. In figure 10 we show the pulse height spectrum of the 662 keV gamma ray
of 137Cs recorded by means of a LaCl3:Ce crystal. The energy resolution is 3.1% FWHM. The
La K x-ray-escape peak can be well observed. An explanation of the small energy-resolution
values of these new scintillators is most likely to be found in the very small non-proportionality
(section 6.1). Both scintillators are hygroscopic to approximately the same degree as NaI:Tl.
Saint-Gobain Crystals & Detectors has started the development of the new scintillators for
production.
Of the scintillators in table 1 that are of interest for PET, we start the discussion with
BGO (Weber and Monchamp 1973). In this material the Bi3+ ion is the intrinsic luminescence
centre. Compared to e.g. LSO:Ce, the light yield is low and the response time long. The poor
energy and time resolution obtained with BGO in a PET system (section 5) are not completely
due to these less than favourable properties. Firstly it is hardly possible to efficiently collect
all the scintillation light at one end of a scintillator column with a small aspect ratio. Secondly,
simultaneous realization of the best energy resolution and best time resolution is a conflicting
condition. Energy resolution requires a long integration time (charge integrating amplifier), for
time resolution we need fast shaping. Yet, more light will lead to a significant improvement.
From the fourth column of table 1 we can deduce that a 30 mm deep BGO crystal has an
interaction probability of 95% for 511 keV quanta. Of this, 38% is due to photoelectric effect
and 57% due to Compton effect. Obviously in a PET system a fraction of the quanta Compton
scattered in a scintillator will escape and interact in neighbouring crystals. This fraction
will increase with the introduction of narrower columns for improving the position resolution
(section 5). Consequently, the fraction detected in the full-energy peak will decrease and, if
the energy selection window is relaxed for compensation, it will result in the opposite to what
we are aiming for: deterioration of the position resolution. The Compton scattering in the
Topical Review
R101
Figure 10. Pulse height spectrum of 662 keV gamma rays detected in LaCl3:10%Ce crystal
(∼3 × 3 × 10 mm3) coupled to a PMT (R1791, shaping time 10 µs). The energy resolution of the
full-energy peak is 3.1% FWHM.
forward direction has the highest probability (angle with direction of the incoming 511 keV
quantum ∼32◦ ). Observation of individual crystal responses and DOI will help to optimize
position resolution.
LSO:Ce was introduced about 10 years ago (Melcher 1990, Melcher and Schweitzer
1992). On one hand the host material was selected for efficient detection (high ρ and Z ),
and on the other the Ce3+ LC ion can be substituted for Lu3+. It appeared to be very difficult
to grow large stress-free crystals (∼1000 cm3) of which small entities can be cut efficiently.
These large crystals were reported to be inhomogeneous in light production and decay time,
and the gamma-ray energy resolution is poorer than expected on the basis of the light yield
(Melcher et al 1999). To some extent the latter is in agreement with the observed strong
non-proportionality (Dorenbos et al 1995).
Using the information of column four in table 1, we deduce that 89% of the 511 keV quanta
interact in the 25 mm LSO:Ce columns used in the ECAT Accel system, to be compared with
95% in 30 mm BGO. Furthermore, of the 89% an amount of 28% is due to the photoelectric
effect and 61% due to the Compton effect. Obviously the process of absorption of a Comptonscattered photon in the same crystal as the Compton electron will contribute to the content of
the full-energy peak. The non-proportionality, however, will result in a significantly poorer
energy resolution.
Samples from different parts of a crystal boule, with different decay times, were
intentionally selected for utilization in the Siemens/CTI HRRT system (section 5), thus
providing DOI by pulse-shape discrimination. Note that the 511 keV interaction efficiency in
7.5 + 7.5 mm of LSO:Ce is only 73%.
Recently, the quality of LSO:Ce boules seems to have improved. LSO is available from
Siemens/CTI (Knoxville, USA). Two other LSO:Ce-related scintillators are also commercially
available: MLS (mixed lutetium silicate, produced by UTAR, Canada) which does not differ
significantly from LSO:Ce, and Lu1.8Gd0.2SiO5:Ce (LGSO, produced by Hitachi), which is
less efficient, has a lower light yield and poorer energy resolution (Pichler et al 1999).
R102
Topical Review
It should be noted that the application of scintillators containing lutetium has two
disadvantages: (a) a high price of ∼$ 50/cm3 due to Lu and (b) presence of the radioactive
isotope 176Lu which gives a count rate of ∼300 s−1 cm−3 (beta decay, end-point energy 565 keV
and some gamma rays in coincidence). For PET (b) is less important.
GSO:Ce was introduced by Hitachi in the early 1980s (Takagi and Fukazawa 1983) and
is still commercially available from this company. The material has a strong analogy with
LSO:Ce. However, in LSO the Ce concentration is maximal at ∼0.06 mol% whereas in
GSO:Ce it can easily be varied. This has an important consequence. The decay time can be
varied as well. It decreases from ∼220 ns to ∼45 ns when the Ce concentration is increased
from ∼0.1 mol% to ∼1 mol% (Ishibashi et al 1989). This phenomenon is explained by
the presence of a Gd sublattice that plays an important role in transporting the available
energy to the Ce ions (Dorenbos et al 1997). The light output is maximal at ∼0.6 mol% and
consequently this is the favourable concentration, resulting in a 60 ns decay time. Originally,
it was difficult to grow large crystals due to the presence of stresses, which caused cracking.
However, improvements in the 1990s resulted in a technique for growing large crystals, e.g.
8 cm diameter × 28 cm length (Ishibashi et al 1997).
The interaction efficiency in the 20 mm long GSO:Ce crystals used in the ADAC/UGM
brain PET system (section 5) is 75%. With LSO:Ce this would be 82%, with BGO 85%.
Furthermore, the photopeak efficiency in GSO:Ce is smaller than that for the other scintillators
and consequently it is to be expected that the content of the full-energy peak, including
Compton-scattered events, is even smaller in relative measure than the total interaction
numbers. The gain in energy and time resolution are clearly at the expense of efficiency.
The scintillator YAP:Ce is clearly only of interest for small-animal PET systems and we
will not discuss it here in detail. The scintillator is commercially available from Saint-Gobain
Crystals & Detectors and from Crytur.
LuAP:Ce is a very interesting candidate to replace BGO in PET. Its light yield is higher
and the response time is much faster than those of BGO. Consequently, an excellent time
resolution is expected. The energy resolution obtained so far is relatively poor. This is mainly
due to the strong scintillation-light absorption, which is still an open problem. Though the
probability of photoelectric effect is smaller, i.e. that of the Compton effect is higher, the
attenuation length is almost identical to that of BGO. The application of LuAP:Ce in detectors
with combinations of different crystals for DOI may also become interesting.
LuAP:Ce was first proposed as a scintillator in 1994 (Minkov 1994) and more detailed
papers appeared shortly after (Moses et al 1995, Lempicki et al 2001). It is difficult to grow
LuAP scintillation crystals. LuAP is metastable at temperatures close to the melting point.
One ends up easily with Lu3Al5O12(LuAG). Yet, several groups, e.g. Crytur Ltd (Turnov,
Czech Republic) and A G Petrosyan (Armenian National Academy of Science), were able to
supply LuAP:Ce crystals for research (Mares et al 1997, Dujardin et al 1998). Some groups
tried to introduce improvements and facilitate the crystal-growing process by adding Gd or
Y (Mares et al 1997). Light yields of 1–2 × 104 photons/MeV are reported for these mixed
crystals, but in general longer (∼100 ns) decay time components are introduced. Research is
in full progress, particularly by the Crystal Clear Collaboration at CERN.
Another relatively new scintillator with a high light yield and fast response is LPS:Ce
(lutetium pyro silicate) (Pauwels et al 2000). The attenuation length of this scintillator is equal
to that of GSO but the photopeak efficiency is higher. Consequently, LPS may be of interest
for a PET system with multi-crystal detectors for DOI information. So far, only small samples
of LPS have been grown. Research on this material is in progress.
From the above discussion it is clear that new scintillators have been introduced or are
being developed for application in PET. It is also clear that improvements concerning energy,
Topical Review
R103
time and position resolution are at the expense of efficiency. This explains why much research
is going on. As already mentioned, Lu2O3- and BaHfO3-based ceramic scintillators are, in
principle, of interest for application in PET provided that better results can be realized on
light yield, fast response time and energy resolution. In addition, material properties such as
machinability should be favourable. Another direction, which is presently being pursued, is
the upgrading of PbWO4 (PWO). This scintillator was extensively studied and developed for
application in high-energy physics, namely, in an electromagnetic calorimeter at CERN. The
4
= 268 × 106 (see table 1 for comparison). Furthermore, the
density is 8.2 g cm−3 and ρZeff
response time is fast (15 ns). However, the light yield is only 200 photons/MeV. Programmes
are aiming at improving the latter by appropriate doping, without increasing the response time.
So far, efforts resulted in a gain in light yield by a factor of ∼3. Obviously, there is a long way
to go. For more information on PWO, see e.g. Mikhailin (2000).
7. Conclusions
For all medical imaging systems using x-rays or gamma rays, radiation detector development in
general and scintillator development in particular are in full progress. There is a strong interest
in the introduction of new dense, high-atomic-number inorganic scintillation crystals with a
high light yield and a fast response, especially for PET. Ceramic scintillators are of interest
for both CT and PET. For PET, research is focused on Ce3+ doped scintillators, employing the
5d → 4f transitions. A high light yield is expected in the visible region. The time response
will be in the 25–100 ns range. Improved energy resolution will also be a point of interest. For
CT, time response requirements are at the microsecond level for decay time and the afterglow
should be well below 10−4 after 3 ms. For x-ray screens light spreading should be kept under
control, e.g. by columnar growth, and afterglow and hysteresis effects should be minimal.
Acknowledgments
The author is indebted to Dr V R Bom (Delft University), Dr J Jansen (Delft University),
Dr C L Melcher (CTI Knoxville), G Muehllehner (ADAC/UGM Philadelphia), P F van
der Stelt (Academic Center for Dentistry, Amsterdam), H Wieckzorec (Philips Research
Laboratories, Aachen) and J Zoetelief (Delft University) for supplying valuable information.
References
Adam L-E, Karp J S and Daube-Witherspoon M E 2001 Evaluation of performance of the CPET scanner using
standardized measurement techniques 2000 IEEE NSS/MIC Conference Record CDROM 17 46–50
Beyer T, Townsend D W, Brun T, Kinahan P E, Charron M, Roddy R, Jerin J, Young J, Byars L and Nutt R 2000 A
combined PET/CT scanner for clinical oncology J. Nucl. Med. 41 1369–79
Blasse G and Grabmaier B C 1994 Luminescent Materials (Berlin: Springer)
Bourgeois Ph, Fanet H, Faure J L, Goret Ph Ille B, El Mamouni H, Peigneux J P and Pouthas J (eds) 1999 Proceedings
of the 2nd international conference on new developments in photon detection Nucl. Instrum. Methods A 442
Casey M E and Nutt R 1986 A multicrystal two dimensional BGO detector system for positron emission tomography
IEEE Trans. Nucl. Sci. 33 460–63
Chepel V, Lopes M I, Kuchenkov A, Ferreira M R and Policarpo A J P L 1997 Performance study of liquid xenon
detector for PET Nucl. Instrum. Methods A 392 427–32
Cherry S R, Dahlbom M and Hoffman E J 1991 3D PET using a conventional multislice tomograph without
septa J. Comput. Assist. Tomogr. 15 655–68
R104
Topical Review
Collot J, Jan S and Tournefier E 2000 A liquid xenon PET camera for neuro-science Proc. 9th Int. Conf. on Calorimetry
in High Energy Physics CALOR 2000 (Frascati Phys. Ser. 21) ed B Aubert, J Colas, P Nédélec and L Poggioli
pp 305–13
De Grado T R, Turkington T G, Williams J J, Stearns C W, Hoffman J M and Coleman R E 1994 Performance
characteristics of a whole-body PET scanner J. Nucl. Med. 35 1398–406
Del Guerra A, Damiani C, Di Domenico G, Motta A, Giganti M, Marchesini R, Piffanelli A, Sabba N, Sartori L and
Zavattini G 2000 An integrated PET-SPECT small animal imager: preliminary results IEEE Trans. Nucl. Sci.
47 1537–40
Dorenbos P, de Haas J T M and van Eijk C W E 1995 Non-proportionality in the scintillation response and the energy
resolution obtainable with scintillation crystals IEEE Trans. Nucl. Sci. 42 2190–202
Dorenbos P and van Eijk C W E (eds) 1996 Proc. Int. Conf. on Inorganic Scintillators and their Applications SCINT95
(Delft: Delft University Press)
Dorenbos P, van’t Spijker J C and van Eijk C W E 1997 Luminescence and scintillation mechanisms in Ce3+ activated
Gd-compounds Proc. Int. Conf. on Inorganic Scintillators and their Applications SCINT97 (Shanghai: Shanghai
Branch Press) pp 307–10
Dujardin C, Garcia-Murillo A, Pédrini C, Madej C, Goutaudier C, Koch A, Petrosyan A G, Ovanesyan K L,
Shirinyan G O and Weber M J 2000 Synthesis and scintillation properties of some dende x-ray phosphors Proc.
5th Int. Conf. on Inorganic Scintillators and their Applications SCINT99 (Moscow: M V Lomonosov Moscow
State University) pp 527–31
Dujardin C et al 1998 Optical and scintillation properties of large LuAlO3:Ce3+ J. Phys.: Condens. Matter 10 3061–73
Garnier N, Dujardin C, Belsky A N C, Moy J P, Wieckzorek H, Chevallier P and Firsov A 2000 Spectroscopy
of CsI(Tl) Proc. 5th Int. Conf. on Inorganic Scintillators and their Applications SCINT99 (Moscow: M V
Lomonosov Moscow State University) pp 379–84
Grabmaier B C and Rossner W 1993 New scintillators for x-ray computed tomography Nucl. Tracks Radiat. Meas.
21 43–5
Gray J E, Stears J G, Swensen S J and Bunch P C 1993 Evaluation of resolution and sensitometric characteristics of
an asymmetric screen-film imaging system Radiology 188 537–39
Greskovich C and Duclos S 1997 Ceramic scintillators Ann. Rev. Mater. Sci. 27 69–88
Happek U, Basun S A, Choi J, Krebs J K and Raukas M 2000 Electron transfer processes in rare earth doped insulators
J. Alloys Compounds 303–304 198–206
Hell E, Knupfer W and Mattern D 2000 The evolution of scintillating medical detectors Nucl. Instrum. Methods A
454 40–8
Holl I, Lorenz E and Mageras G 1988 A measurement of the light yield of common inorganic scintillators IEEE
Trans. Nucl. Sci. 35 105–9
Hupke R and Doubrava C 1999 The new UFC-detector for CT-imaging Phys. Medica XV 315–8
Ishibashi H, Kurashige K, Kurata Y, Susa K, Kurosawa T, Kobayashi M, Tanaka M, Hara K and Ishii M 1997 Large
Ce-doped Gd2SiO5 (GSO) single crystals 80 mm in diameter and its performance Proc. Int. Conf. on Inorganic
Scintillators and their Applications SCINT97 (Shanghai: Shanghai Branch Press) pp 295–98
Ishibashi H, Shimizu K, Susa K and Kubota S 1989 Cerium doped GSO scintillators and its application to position
sensitive detectors IEEE Trans. Nucl. Sci. 36 170–2
Jasczak R J, Coleman R E and Lim C B 1980 SPECT: single photon emission computed tomography IEEE Trans.
Nucl. Sci. 27 1137–53
Jing T, Goodman C A, Drewery J, Cho G, Hong W S, Lee H, Kaplan S N, Mireshghi A, Perez-Mendez V and
Wildermuth D 1994 Amorphous silicon pixel layers with cesium iodide converters for medical radiography
IEEE Trans. Nucl. Sci. 41 903–9
Jones W F, Casey M E, van Lingen A and Bendriem B 2001 LSO PET/SPECT spatial resolution: critical on-line
DOI rebinning methods and results 2000 IEEE NSS/MIC Conference Record CDROM 16 54–8
Kalender W A 2000 Computed Tomography (Munich: MCD)
Karp J S and Daube-Witherspoon M E 1987 Depth-of-interaction determination in NaI(Tl) and BGO scintillation
crystals using a temperature gradient Nucl. Instrum. Methods A 260 509–17
Karp J S, Surti S, Freifelder R, Daube-Witherspoon M E, Cardi C, Adam L-E, Chase B, Vaska P and Muehllehner G
2001 Performance of a GSO brain PET camera 2000 IEEE NSS/MIC Conference Record CDROM 17 7–11
Lakshmanan A R, Murase N, Yazawa T, Qiu J, Mitsuyu T, Hirao K, Tomoto A and Hoffmann W 2001 Luminescence
studies in BaFBr and BaFBr:Eu Radiat. Meas. 33 119–27
Lecoq P 2001 Clear-PEM, a dedicated PET camera for mammography 6th Int. Conf. on Inorganic Scintillators and
their use in Scientific and Industrial Applications SCINT2001 (Chamonix, France 16–21 Sept. 2001)
Topical Review
R105
Lempicki A, Brecher C, Lingertat H, Szupryczynski P, Nagarkar V V, Tipnis S V and Miller S R 2001 A new
scintillator for digital x-ray radiography Book of abstracts of the 6th Int. Conf. on Inorganic Scintillators and
Their use in Scientific and Industrial Applications SCINT2001 (Chamonix, France, 16–21 Sept. 2001) MI-O 02
Lempicki A, Randles M H, Wisniewski D, Balcerzyk M, Brecher C and Wojtowicz A 1995 LuAlO3:Ce and other
aluminate scintillators IEEE Trans. Nucl. Sci. 42 280–84
Mares J A, Nikl M, Mihokova E, Kvapil J, Giba J and Blazek K 1997 Spectroscopy and transfer processes in
LuxGd1−xAlO3:Ce scintillators J. Lumin. 72–74 737–9
McCollough C H and Zink F E 1999 Performance evaluation of a multi-slice CT system Med. Phys. 26 2223–30
Melcher C L 1990 US Patent No 4958080
Melcher C L, Schmand M, Eriksson M, Eriksson L, Casey M, Nutt R, Lefaucheur J L and Chai B 1999 Scintillation
properties of LSO:Ce Boules 1998 IEEE Nuclear Science Symp. & Medical Imaging Conf. Record (Toronto,
Ontario, Canada 8–14 Nov. 1998) (CDROM) N9-3
Melcher C L and Schweitzer J S 1992 Cerium-doped lutetium oxyorthosilicate: a fast, efficient new scintillator IEEE
Trans. Nucl. Sci. 39 502–5
Meng L J and Ramsden D 2000 Performance results of a prototype depth-encoding PET detector IEEE Trans. Nucl.
Sci. 47 1011–17
Mikhailin V (ed) 2000 Proc. 5th Int. Conf. on Inorganic Scintillators and their Applications SCINT99 (Moscow:
M V Lomonosov Moscow State University)
Minkov B I 1994 Promising new lutetium based single crystals for fast scintillation Functional Materials 1 103–05
Miura N 1998 Phosphors for x-ray and ionizing radiation Phosphor Handbook ed S Shionoya and W M Yen (Boca
Raton, FL: CRC Press) ch 7.1 pp 521–30
Moses W W, Derenzo S E, Fyodorov A, Korzhik M, Gektin A, Minkov B and Aslanov V 1995 LuAlO3:Ce—a high
density, high speed scintillator for gamma detection IEEE Trans. Nucl. Sci. 42 275–9
Moy J-P 1999 Recent developments in x-ray imaging detectors Nucl. Instrum. Methods A 442 26–37
Muehllehner G, Karp J S and Surti S 2001 Design considerations for PET scanners Q. J. Nucl. Med. at press
Muller S 1999 Full-field digital mammography designed as a complete system Eur. J. Radiol. 31 25–34
Nakamura R and Ishii M 2000 Improvements in the x-ray characteristics of Gd2O2S:Pr ceramic Scintillators Proc. 5th
Int. Conf. Inorganic Scintillators and their Applications SCINT99 501–05 (Moscow: M V Lomonosov Moscow
State University)
Neitzel U, Maack I and Günther-Hohfahl S 1994 Image quality of a digital chest radiography system based on a
selenium detector Med. Phys. 21 509–16
NSS/MIC 1999 Nuclear science symposium & medical imaging conference IEEE Trans. Nucl. Sci. 46
NSS/MIC 2000 Nuclear science symposium & medical imaging conference IEEE Trans. Nucl. Sci. 47
Pauwels D, Le Masson N, Viana B, Kahn-Harari A, van Loef E V D, Dorenbos P and van Eijk C W E 2000 A novel
inorganic scintillator: Lu2Si2O7:Ce3+ (LPS) IEEE Trans. Nucl. Sci. 47 1787–90
Pichler B J, Böning G, Rafecas M, Schlosshauer M, Lorenz E and Ziegler S I 1999 LGSO scintillation crystals
coupled to new large area APDs compared to LSO and BGO IEEE Trans. Nucl. Sci. 46 289–91
Rodnyi P A, Dorenbos P and van Eijk C W E 1995 Energy loss in inorganic scintillators Phys. Status Solidi b 187
15–29
Rossner W, Ostertag M and Jermann F 1999 Properties and applications of gadolinium oxysulfide based ceramic
scintillators Electrochem. Soc. Proc. 98-24 187–94
Saoudi A and Lecomte R 1999 A novel APD-based detector module for multi-modality PET/SPECT/CT scanners
IEEE Trans. Nucl. Sci. 46 479–84
Saoudi A, Pepin C M, Dion F, Bentourkia M, Lecomte R, Andreaco M, Casey M, Nutt R and Daudet H 1999
Investigation of depth-of-interaction by pulse shape discrimination in multicrystal detectors read out by avalanche
photodiodes IEEE Trans. Nucl. Sci. 46 462–67
Schiebel U, Conrads N, Jung N, Weibrecht M, Wieczorek H and Zaengel T 1994 Fluoroscopic x-ray imaging with
amorphous silicon thin-film arrays SPIE Phys. Med. Imaging 2163 129–40
Seidel J, Vaquero J J, Siegel S, Gandler W R and Green M V 1999 Depth identification accuracy of a three layer
Phoswich PET detector module IEEE Trans. Nucl. Sci. 46 485–90
Shah K S, Farrell R, Grazioso R, Shao Y and Cherry S R 2001 APD arrays for medical imaging, 6th Int. Conf. on
Inorganic Scintillators and their use in Scientific and Industrial Applications SCINT2001 (Chamonix, France,
16–21 Sept. 2001)
Short M D 1984 Gamma camera systems Nucl. Instrum. Methods 221 142–9
Soltani P K, Wysnewski D and Swartz K 1999 Amorphous selenium direct radiography for industrial imaging Proc.
Computerized Tomography and Image Processing (Berlin: DGZfP Proceedings BB 67-CD) pp 123–32
Sonoda M, Takano M, Miyahara J and Kato H 1983 Computed radiography utilizing scanning laser stimulated
luminescence Radiology 148 833–8
R106
Topical Review
Spaeth J-M 2001 Recent developments in x-ray storage phosphor materials Radiat. Meas. 33 527–32
Takagi K and Fukazawa T 1983 Cerium-activated Gd2SiO5 single crystal scintillator Appl. Phys. Lett. 42 43–5
Townsend D W 2001 New trends in PET scanners presented by C Morel at the 6th Int. Conf. on Inorganic Scintillators
and their use in Scientific and Industrial Applications SCINT2001 (Chamonix, France, 16–21 Sept. 2001)
Townsend D W, Spinks T J, Jones T, Geissbühler A, Defrise M, Gilardi M C and Heather J 1989 Three-dimensional
reconstruction of PET data from a multi-ring camera IEEE Trans. Nucl. Sci. 36 1056–65
Valentine J D and Rooney B D 1997 Intrinsic energy resolution and its dependence on light yield nonproportionality
Proc. Int. Conf. on Inorganic Scintillators and their Applications SCINT97 (Shanghai: Shanghai Branch Press)
pp 91–4
van der Stelt P F 2001 The implementation of digital sensors in maxillofacial radiography Nucl. Instrum. Methods A
460 45–9
van Eijk C W E 2001 Inorganic-scintillator development Nucl. Instrum. Methods A 460 1–4
van Eijk C W E, Dorenbos P, van Loef E V D, Krämer K and Güdel H U 2001 Energy resolution of some new
inorganic-scintillator gamma-ray detectors Radiat. Meas. 33 521–5
van Loef E V D, Dorenbos P, van Eijk C W E, Krämer K and Güdel H U 2000 High-energy-resolution scintillator:
Ce3+ activated LaCl3 Appl. Phys. Lett. 77 1467–8
van Loef E V D, Dorenbos P, van Eijk C W E, Krämer K and Güdel H U 2001 High-energy-resolution scintillator:
Ce3+ activated LaBr3 Appl. Phys. Lett. 79 1573–5
Venkataramani V, Loureiro S, Rane M, Duclos S, Stearns C and McDaniel D 2001 Transparant ceramic routes to
scintillators 2001 IEEE Nuclear Science Symp & Medical Imaging Conf. (San Diego, CA, N8–4, 4–10 Nov.
2001) at press
Webb S (ed) 1990 The Physics of Medical Imaging (Bristol: Adam Hilger)
Weber S, Bauer A, Herzog H, Kehren F, Mühlensiepen H, Vogelbruch J, Coenen H H, Zilles K and Halling H 2000
Recent results of the TierPET Scanner IEEE Trans. Nucl. Sci. 47 1665–9
Weber M J and Monchamp R R 1973 Luminescence of Bi4Ge3O12: spectral and decay properties J. Appl. Phys. 44
5495–9
Wieckzorec H and Overdick M 2000 Afterglow and hysteresis in CsI:Tl Proc. 5th Int. Conf. on Inorganic Scintillators
and their Applications (Moscow: M L Lomonosov Moscow State University) pp 385–90
Wienhard K, Dahlbom M, Eriksson L, Michel C, Bruckbauer T, Pietrzyk U and Heiss W-D 1994 The ECAT EXACT
HR: performance of a new high resolution positron scanner J. Comput. Assist. Tomogr. 18 110–8
Wienhard K et al 2001 The ECAT HRRT: performance and first clinical application of the new high resolution
research tomograph 2000 IEEE NSS/MIC Conference Record CDROM 17 2–6
Williams J R and Thwaites D I 2000 Radiotherapy Physics in Practice (Oxford: Oxford University Press)
Yamada H, Suzuki A, Uchida Y, Yoschida M and Yamamoto H 1989 A scintillator Gd2O2S:Pr,Ce,F for x-ray computed
tomography J. Electrochem. Soc. 136 2713–20
Yin Zhiwen, Feng Xiqi, Li Peijun and Xue Zhilin (eds) 1997 Proc. Int. Conf. on Inorganic Scintillators and their
Applications SCINT97 (Shanghai: Shanghai Institute of Ceramics, Chinese Academy of Sciences, Shanghai
Branch Press)